[prev in list] [next in list] [prev in thread] [next in thread] 

List:       drbd-user
Subject:    Re: [DRBD-user] Extremely high latency problem
From:       Bret Mette <bret.mette () dbihosting ! com>
Date:       2014-06-05 23:00:20
Message-ID: CAKpThoxjeihALooYQJGeDLDPxWDDDguboaSO3a27NCxvDbFtYQ () mail ! gmail ! com
[Download RAW message or body]

[Attachment #2 (multipart/alternative)]


Hey Phillip,

I upgraded to 8.4.4 and it totally resolved my issue. My targets went from
0.5 - 1.1M/sec writes to 30M/sec writes on random writes and 7M/sec (with
caching turned on) to 85M/sec (with no caching!).


Thank you so much, I couldn't be happier.


- Bret


On Thu, Jun 5, 2014 at 3:34 AM, Philipp Marek <philipp.marek@linbit.com>
wrote:

> Hi Bret,
>
> > Intel(R) Xeon(R) CPU E3-1230 V2 @ 3.30GH
> > CentoS 6.5 - 2.6.32-431.17.1.el6.x86_64
> > drbd version: 8.3.16 (api:88/proto:86-97)
> please use 8.4.4 (or .5).
> Eg because of http://blogs.linbit.com/p/469/843-random-writes-faster/
>
>
> > dd if=/dev/zero of=./testbin  bs=512 count=1000 oflag=direct
> > 512000 bytes (512 kB) copied, 32.3254 s, 15.8 kB/s
> This means ~32msec per write.
>
> But: you're writing into a _filesystem_, not directly onto the DRBD
> resource.
> And that means quite some write amplification - each 512 byte write
> _might_ result (depending on the filesystem) in _multiple_ write-
> requests, each with a separate barrier ...
>
>
> Please try writing into a DRBD device, without any filesystem inbetween;
> and using 4kB blocks might help to avoid read/modify/write cycles, too.
>
>
>

[Attachment #5 (text/html)]

<div dir="ltr">Hey Phillip,<div><br></div><div>I upgraded to 8.4.4 and it totally \
resolved my issue. My targets went from 0.5 - 1.1M/sec writes to 30M/sec writes on \
random writes and 7M/sec (with caching turned on) to 85M/sec (with no \
caching!).</div> <div><br></div><div><br></div><div>Thank you so much, I couldn&#39;t \
be happier.</div><div><br></div><div><br></div><div>- Bret</div></div><div \
class="gmail_extra"><br><br><div class="gmail_quote">On Thu, Jun 5, 2014 at 3:34 AM, \
Philipp Marek <span dir="ltr">&lt;<a href="mailto:philipp.marek@linbit.com" \
target="_blank">philipp.marek@linbit.com</a>&gt;</span> wrote:<br> <blockquote \
class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc \
solid;padding-left:1ex">Hi Bret,<br> <div class=""><br>
&gt; Intel(R) Xeon(R) CPU E3-1230 V2 @ 3.30GH<br>
&gt; CentoS 6.5 - 2.6.32-431.17.1.el6.x86_64<br>
&gt; drbd version: 8.3.16 (api:88/proto:86-97)<br>
</div>please use 8.4.4 (or .5).<br>
Eg because of <a href="http://blogs.linbit.com/p/469/843-random-writes-faster/" \
target="_blank">http://blogs.linbit.com/p/469/843-random-writes-faster/</a><br> <div \
class=""><br> <br>
&gt; dd if=/dev/zero of=./testbin   bs=512 count=1000 oflag=direct<br>
&gt; 512000 bytes (512 kB) copied, 32.3254 s, 15.8 kB/s<br>
</div>This means ~32msec per write.<br>
<br>
But: you&#39;re writing into a _filesystem_, not directly onto the DRBD<br>
resource.<br>
And that means quite some write amplification - each 512 byte write<br>
_might_ result (depending on the filesystem) in _multiple_ write-<br>
requests, each with a separate barrier ...<br>
<br>
<br>
Please try writing into a DRBD device, without any filesystem inbetween;<br>
and using 4kB blocks might help to avoid read/modify/write cycles, too.<br>
<br>
<br>
</blockquote></div><br></div>



_______________________________________________
drbd-user mailing list
drbd-user@lists.linbit.com
http://lists.linbit.com/mailman/listinfo/drbd-user


[prev in list] [next in list] [prev in thread] [next in thread]